跳转至

Theano 条件语句

theano 中提供了两种条件语句,ifelseswitch,两者都是用于在符号变量上使用条件语句:

  • ifelse(condition, var1, var2)
    • 如果 conditiontrue,返回 var1,否则返回 var2
  • switch(tensor, var1, var2)
    • Elementwise ifelse 操作,更一般化
  • switch 会计算两个输出,而 ifelse 只会根据给定的条件,计算相应的输出。

ifelse 需要从 theano.ifelse 中导入,而 switchtheano.tensor 模块中。

In [1]:

import theano, time
import theano.tensor as T
import numpy as np
from theano.ifelse import ifelse
Using gpu device 1: Tesla K10.G2.8GB (CNMeM is disabled)

假设我们有两个标量参数:\(a, b\),和两个矩阵 \(\mathbf{x, y}\),定义函数为:

\[ \mathbf z = f(a, b,\mathbf{x, y}) = \left\{ \begin{aligned} \mathbf x & ,\ a <= b\\="" \mathbf="" y="" &="" ,\="" a=""> b \end{aligned} \right. \]

定义变量:

In [2]:

a, b = T.scalars('a', 'b')
x, y = T.matrices('x', 'y')

ifelse 构造,小于等于用 T.lt(),大于等于用 T.gt()

In [3]:

z_ifelse = ifelse(T.lt(a, b), x, y)

f_ifelse = theano.function([a, b, x, y], z_ifelse)

switch 构造:

In [4]:

z_switch = T.switch(T.lt(a, b), x, y)

f_switch = theano.function([a, b, x, y], z_switch)

测试数据:

In [5]:

val1 = 0.
val2 = 1.
big_mat1 = np.ones((10000, 1000), dtype=theano.config.floatX)
big_mat2 = np.ones((10000, 1000), dtype=theano.config.floatX)

比较两者的运行速度:

In [6]:

n_times = 10

tic = time.clock()
for i in xrange(n_times):
    f_switch(val1, val2, big_mat1, big_mat2)
print 'time spent evaluating both values %f sec' % (time.clock() - tic)

tic = time.clock()
for i in xrange(n_times):
    f_ifelse(val1, val2, big_mat1, big_mat2)
print 'time spent evaluating one value %f sec' % (time.clock() - tic)
time spent evaluating both values 0.638598 sec
time spent evaluating one value 0.461249 sec


回到顶部